From: Jianjun Duan <duanj@linux.vnet.ibm.com>
To: David Gibson <david@gibson.dropbear.id.au>,
"Dr. David Alan Gilbert (git)" <dgilbert@redhat.com>
Cc: qemu-devel@nongnu.org, amit.shah@redhat.com, quintela@redhat.com,
pasic@linux.vnet.ibm.com
Subject: Re: [Qemu-devel] [very-WIP 1/7] migration: Add VMSTATE_WITH_TMP
Date: Mon, 17 Oct 2016 11:49:02 -0700 [thread overview]
Message-ID: <61d82189-401b-7063-ffa5-4442c7d11d25@linux.vnet.ibm.com> (raw)
In-Reply-To: <20161017033134.GM25390@umbus.fritz.box>
On 10/16/2016 08:31 PM, David Gibson wrote:
> On Tue, Oct 11, 2016 at 06:18:30PM +0100, Dr. David Alan Gilbert (git) wrote:
>> From: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
>>
>> VMSTATE_WITH_TMP is for handling structures where some calculation
>> or rearrangement of the data needs to be performed before the data
>> hits the wire.
>> For example, where the value on the wire is an offset from a
>> non-migrated base, but the data in the structure is the actual pointer.
>>
>> To use it, a temporary type is created and a vmsd used on that type.
>> The first element of the type must be 'parent' a pointer back to the
>> type of the main structure. VMSTATE_WITH_TMP takes care of allocating
>> and freeing the temporary before running the child vmsd.
>>
>> The post_load/pre_save on the child vmsd can copy things from the parent
>> to the temporary using the parent pointer and do any other calculations
>> needed; it can then use normal VMSD entries to do the actual data
>> storage without having to fiddle around with qemu_get_*/qemu_put_*
>>
If customized put/get can do transformation and dumping/loading data
to/from the parent structure, you don't have to go through
pre_save/post_load, and may get rid of parent pointer.
Thanks,
Jianjun
>> Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
>
> The requirement for the parent pointer is a little clunky, but I don't
> quickly see a better way, and it is compile-time verified. As noted
> elsewhere I think this is a really useful approach which could allow a
> bunch of internal state cleanups while preserving migration.
>
> Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
>
>> ---
>> include/migration/vmstate.h | 20 ++++++++++++++++++++
>> migration/vmstate.c | 38 ++++++++++++++++++++++++++++++++++++++
>> 2 files changed, 58 insertions(+)
>>
>> diff --git a/include/migration/vmstate.h b/include/migration/vmstate.h
>> index 9500da1..efb0e90 100644
>> --- a/include/migration/vmstate.h
>> +++ b/include/migration/vmstate.h
>> @@ -259,6 +259,7 @@ extern const VMStateInfo vmstate_info_cpudouble;
>> extern const VMStateInfo vmstate_info_timer;
>> extern const VMStateInfo vmstate_info_buffer;
>> extern const VMStateInfo vmstate_info_unused_buffer;
>> +extern const VMStateInfo vmstate_info_tmp;
>> extern const VMStateInfo vmstate_info_bitmap;
>> extern const VMStateInfo vmstate_info_qtailq;
>>
>> @@ -651,6 +652,25 @@ extern const VMStateInfo vmstate_info_qtailq;
>> .offset = offsetof(_state, _field), \
>> }
>>
>> +/* Allocate a temporary of type 'tmp_type', set tmp->parent to _state
>> + * and execute the vmsd on the temporary. Note that we're working with
>> + * the whole of _state here, not a field within it.
>> + * We compile time check that:
>> + * That _tmp_type contains a 'parent' member that's a pointer to the
>> + * '_state' type
>> + * That the pointer is right at the start of _tmp_type.
>> + */
>> +#define VMSTATE_WITH_TMP(_state, _tmp_type, _vmsd) { \
>> + .name = "tmp", \
>> + .size = sizeof(_tmp_type) + \
>> + QEMU_BUILD_BUG_EXPR(offsetof(_tmp_type, parent) != 0) + \
>> + type_check_pointer(_state, \
>> + typeof_field(_tmp_type, parent)), \
>> + .vmsd = &(_vmsd), \
>> + .info = &vmstate_info_tmp, \
>> + .flags = VMS_LINKED, \
>> +}
>> +
>> #define VMSTATE_UNUSED_BUFFER(_test, _version, _size) { \
>> .name = "unused", \
>> .field_exists = (_test), \
>> diff --git a/migration/vmstate.c b/migration/vmstate.c
>> index 2157997..f2563c5 100644
>> --- a/migration/vmstate.c
>> +++ b/migration/vmstate.c
>> @@ -925,6 +925,44 @@ const VMStateInfo vmstate_info_unused_buffer = {
>> .put = put_unused_buffer,
>> };
>>
>> +/* vmstate_info_tmp, see VMSTATE_WITH_TMP, the idea is that we allocate
>> + * a temporary buffer and the pre_load/pre_save methods in the child vmsd
>> + * copy stuff from the parent into the child and do calculations to fill
>> + * in fields that don't really exist in the parent but need to be in the
>> + * stream.
>> + */
>> +static int get_tmp(QEMUFile *f, void *pv, size_t size, VMStateField *field)
>> +{
>> + int ret;
>> + const VMStateDescription *vmsd = field->vmsd;
>> + int version_id = field->version_id;
>> + void *tmp = g_malloc(size);
>> +
>> + /* Writes the parent field which is at the start of the tmp */
>> + *(void **)tmp = pv;
>> + ret = vmstate_load_state(f, vmsd, tmp, version_id);
>> + g_free(tmp);
>> + return ret;
>> +}
>> +
>> +static void put_tmp(QEMUFile *f, void *pv, size_t size, VMStateField *field,
>> + QJSON *vmdesc)
>> +{
>> + const VMStateDescription *vmsd = field->vmsd;
>> + void *tmp = g_malloc(size);
>> +
>> + /* Writes the parent field which is at the start of the tmp */
>> + *(void **)tmp = pv;
>> + vmstate_save_state(f, vmsd, tmp, vmdesc);
>> + g_free(tmp);
>> +}
>> +
>> +const VMStateInfo vmstate_info_tmp = {
>> + .name = "tmp",
>> + .get = get_tmp,
>> + .put = put_tmp,
>> +};
>> +
>> /* bitmaps (as defined by bitmap.h). Note that size here is the size
>> * of the bitmap in bits. The on-the-wire format of a bitmap is 64
>> * bit words with the bits in big endian order. The in-memory format
>
next prev parent reply other threads:[~2016-10-17 18:49 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-10-11 17:18 [Qemu-devel] [very-WIP 0/4] Migration: VMSTATE_WITH_TMP Dr. David Alan Gilbert (git)
2016-10-11 17:18 ` [Qemu-devel] [very-WIP 1/7] migration: Add VMSTATE_WITH_TMP Dr. David Alan Gilbert (git)
2016-10-17 3:31 ` David Gibson
2016-10-17 18:49 ` Jianjun Duan [this message]
2016-10-17 18:52 ` Dr. David Alan Gilbert
2016-10-17 19:02 ` Jianjun Duan
2016-10-17 19:16 ` Dr. David Alan Gilbert
2016-10-17 19:30 ` Jianjun Duan
2016-10-18 8:06 ` Dr. David Alan Gilbert
2016-10-11 17:18 ` [Qemu-devel] [very-WIP 2/7] tests/migration: Add test for VMSTATE_WITH_TMP Dr. David Alan Gilbert (git)
2016-10-17 3:34 ` David Gibson
2016-10-11 17:18 ` [Qemu-devel] [very-WIP 3/4] slirp: VMStatify sbuf Dr. David Alan Gilbert (git)
2016-10-17 3:36 ` David Gibson
2016-10-17 17:54 ` Halil Pasic
2016-10-17 19:06 ` Dr. David Alan Gilbert
2016-10-18 10:40 ` Halil Pasic
2016-10-11 17:18 ` [Qemu-devel] [very-WIP 4/4] virtio/migration: Migrate virtio-net to VMState Dr. David Alan Gilbert (git)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=61d82189-401b-7063-ffa5-4442c7d11d25@linux.vnet.ibm.com \
--to=duanj@linux.vnet.ibm.com \
--cc=amit.shah@redhat.com \
--cc=david@gibson.dropbear.id.au \
--cc=dgilbert@redhat.com \
--cc=pasic@linux.vnet.ibm.com \
--cc=qemu-devel@nongnu.org \
--cc=quintela@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).